1. Field of the Invention
The present invention relates to broadband switching networks using ATM (Asynchronous Transfer Mode) technics.
2. Description of the Related Art
Efforts for integrating individual service networks such as telephone networks, data networks, FAX networks, and so forth which have been developed and constructed over 100 years of history into one network system with ISDN (Integrated Services Digital Network) have been made throughout the world.
As the first step for constructing the ISDN system, narrow band ISDN systems have been operated in advanced countries including Japan since 1988. In addition, besides integration with a broadcasting network by using a broadband ISDN based on the ATM technics, the engineering developments of the ISDN network have been initiated by CCITT (International Telegraph and Telephone Consultative Committee) and promoted in major laboratories in the world.
The broadband ISDN network is provided with an ultra high speed user-network interface with a transmission speed of 155.52 Mbps or 622.08 Mbps. Thus, through the same interface, the conventional telephones, facsimile machines, and so forth can be treated as a constant speed service CBR (Continuous Bit Rate), while computer data with large capacity and ultra high speed including motion pictures, such as, high definition TV pictures, CAD (Computer Aided Design) data, and computer graphics data, and so forth can be treated as a variable speed service VBR (Variable Bit Rate). Thus, with the CBR and the VBR services, various data can be flexibly transmitted through the same interface.
However, thus far, before making a communication, the user had to declare call attribute data such as peak traffic, mean traffic, burstiness, terminal equipment type, service quality (for example, cell loss rate, cell transmission delay), and the like upon the network. In accordance with the call attribute data that the user has declared, the network estimates a required communication resource necessary for making the communication with respect to the call, checks the use state of the resource in the network, and determines whether or not to accept the call request. As the result of the determination, when the call request is accepted, the information to be transmitted is divided into packets which have a constant length (53 octets) (the packets are named cells) and then sent to the network. However, occasionally, cells which do not conform with the attribute data being declared may be sent to the network.
Thus, when unexpectedly excessive cells are sent to the network and they are concentrated in one path (as the burstiness is strong, this tendency becomes remarkable), they reside in the network. To prevent that, ATM switches, each of which is a constructional key element of the broadband ISDN network, are provided with a cell buffer with a large storage capacity. However, if such a cell buffer cannot store the cells which stay in the network, they will be lost. This situation is named a cell discard. When the network receives cells and then marks those which exceeds the range of attribute data declared by so-called polling function as violation cells, they are discarded. In addition, when a terminal equipment sends cells as non-priority cells in VBR such as class B (variable bit rate picture communication) (for example, in hierarchical picture coding system, a method where cells are divided into priority cells and non-priority cells depending on their importance is being considered), they are discarded. If the buffer does not fully store the cells even after the marked cells are discarded, cells in class A (circuit emulation communication) or the like will be also discarded.
Generally, a bit error due to noise or the like over the transmission path is checked with a CRC code disposed at the last position of information to be transmitted. When necessary, by issuing a retransmission request to the sender side, the information with respect to a bit error can be restored. However, when the cell discard is performed, since the receiver side cannot know the transmission of cells, it cannot request the retransmission of the cells to the sender side.
Since the cell discard will become a critical problem in data communication with respect to class C (connection oriented) and class D (support of transmission of connection-less data), a sequence number is provided for the information field of each cell (48 octets) as an ATM adaptation layer function. In addition, a mechanism for detecting the cell discard and for issuing a retransmission request on the receiver side is additionally provided.
On the other hand, for calls in the classes A and B, which should be transmitted in real time, CCITT has recommended a coding system which can withstand the cell discard.
Major problems with respect to the broadband ISDN switching networks which have been studied mainly by CCITT are summarized as follows.
(1) Service quality
(a) Cell discard
As was described earlier, by assigning a sequence number, it is possible to detect a cell loss on the receiver side. However, since calls in the classes A and B should be transmitted in real time, it is substantially difficult to restore cells which are lost by the retransmission. By the coding method, which is a future study subject, it may be possible to reduce the adverse effect of the discard of non-priority cells. However, besides the discard of the normal cells (those which are neither violation cells nor non-priority cells), the discard of violation cells will result in critical problems.
In other words, when a honest (innocent) user unconsciously violates the range of the attribute data being declared, cells that the user has transmitted will be discarded regardless of whether they are priority cells or non-priority cells. In other words, the information received by the network may be lost. In addition, the sender side cannot know what and how much information is lost. To prevent that, a prudent and honest user will always have to declare the attribute data with an allowance although he or she knows that the communication fee will becomes expensive.
On the other hand, a user who wants to save the communication fee will declare the attribute data which is rather small while observing the traffic condition of the network although he or she knows that violation cells may take place. In other words, each user will haggle with the network about the negotiation of the attribute data like playing a game therewith. Whenever the user repeats success and failure in the negotiations with the network, his or her action will escalate.
The increase of such users causes the traffic in the network to be abnormally increased and thereby loosing ordinal cells transmitted by prudent and honest users. Thus, the users have suspicion and apprehension about the network. It is inevitable that the essential purpose of the public communication network, which is "correct, fast, and impartial communication transmission", is discarded.
Moreover, even in the classes C and D, the same situation will take place. Particularly, in data communication, as was described above, since the loss of information is never permitted, the retransmission of cells which were lost will be performed in a high rank layer.
As the traffic is heavy, probability of occurrence of a cell discard will become high. When the retransmission of cells which were lost is repeated, the traffic will become much higher. Thus, the network will become congested. In other words, the cell discard will result in deteriorating the stability of the network.
(b) Variation of cell delay
As was described above, in high traffic conditions, cells will reside in the network. In other words, the cells will be transmitted with a delay. As the capacity of the switch (the scale of ATM switch) becomes large and/or the number of relays in the network increases, the amount of delay increases. In addition, the amount of delay varies depending on the traffic condition in the network. This variation of the amount of delay is named the variation of cell delay. For calls in the classes A and B, which should be transmitted in real time, it is necessary to provide a buffer on the receiver terminal equipment side so as to compensate the variation of cell delay. For example, in a relatively small scaled broadband switching network accommodating about 100 lines (interface), the variation of cell delay is in the range from about several .mu. sec to about several msec. However, in an international communication, since there are many relaying networks, the variation of cell delay may be several 100 msec (excluding the absolute delay time involved in long distance transmission). Thus, each terminal equipment should be provided with a buffer with large capacity (several Mbytes for a terminal equipment with an information speed of 100 Mbps). To prevent that, it is necessary to decrease the variation of cell delay itself in the network in a manner to perform priority control in accordance with required service quality by using an ATM switch, which will be described later.
(c) Call connection time
The conventional switches including those for narrow band ISDN systems usually take several seconds, occasionally more than 10 seconds, or 20 seconds for connecting a call (until an originating user of a telephone hears a ring back tone).
When the originating terminal equipment issues a call set request to a switch node, it performs a terminating process for the call. Thereafter, the switch node obtains a line connected from the originating terminal equipment to the terminating terminal equipment and performs an originating process to the terminating terminal equipment.
However, when a call passes through a plurality of relay switch nodes, since each relay switch node individually performs the terminating process, line obtaining process, and originating process, the connection time becomes much longer.
On the other hand, the broadband ISDN network employs the conception of above mentioned logical bus so as to simplify the processes required in each relaying node. However, it is necessary bidirectionaly to negotiate with the network the attribute data which was negotiated upon the set of the call, to obtain a band in a virtual path in accordance with the negotiated result, and to perform an originating process to the terminating terminal equipment (from the originating terminal equipment to the terminating terminal equipment through the network; from the terminating terminal equipment to originating terminal equipment through the network). There is no denying the fact that the call which is connected from the originating terminal equipment to the terminating terminal equipment takes a time on the order of seconds.
However, for calls in the classes A and B, where it is estimated that their communication times are equal to or longer than those of telephones, the connection time on the order of seconds does not result in a remarkable problem.
However, in a computer communication where information is intermittently transmitted, if a call connection takes a time on the order of seconds whenever the information is transmitted, the performance of the computer cannot be satisfactorily used and thereby the operability of the system is degraded. To prevent that, in LAN (Local Area Network), which is a dedicated local area computer communication system, a system named a connection-less is used to allow the user not to realize a long connection time. When the broadband ISDN is practically used, it will be possible to transmit a file with a storage of 1 Mbytes in several msec. Thus, in a computer communication, where the performance is intensively advanced as technologies are rapidly innovated at the present time, it can be said that the utilization of the connection-less system will become mandatory.
On the other hand, the above mentioned class D for "support of transmission of connection-less data" supposes inter-LAN connections. In this class, when a call is initially connected, a connection time equivalent to those in the classes A and B is permitted. After the call is connected, the path is held for a long time. The routing control from end to end is handled by the user (upper layer) rather than the network.
In the LAN system, where the same communication medium is shared by a large number of users, a traffic always takes place over the LAN and between LANS. Thus, even if the path is held in the broadband ISDN network for a long time, it can be operated in commercial basis. However, when one computer terminal equipment is connected directly to the broadband ISDN network, for example in the case where a remote user accesses a central data base, if a path is held in the class D for a long time, it cannot be operated in commercial basis because information is intermittently transmitted as was described earlier. In addition, for example by providing the network with a permanent virtual path, where from the standpoint of the user it seems that a dedicated line is routed between both the ends and thereby a call connection is not required, and with a function for generating the address of a cell header by using the address of an upper header, which are subjects to be studied in future, a connection-less service can be accomplished. However, with the above permanent virtual path and the function, the network should hold a particular communication resource always or until the communication is completed so that it can handle the accessing of each user anytime. Thus, an expensive communication fee will be applied to each user. Most of users who make much account of cost performance along with performance should select the connection oriented data communication in the class C with consideration of connection item on the order of seconds. In other words, it is suggested that the ordinary users frequently issue call connection requests in the network and thereby it tends to be congested and loose the stability thereof.
It is estimated that half the full families will use advanced personal computer communication systems with hypermedia or the like in the year of around 2000. However, unless the broadband ISDN network effectively operates the communication resources at inexpensive cost and provides connection-less services with a light load thereof or reduces a call connection time to the same level as the connection-less services, it will not be able to attract the ordinary computer users which expectedly have huge latent demands (although LAN connection users are limited only to major companies and the like). In addition, with respect to the stability of the network, it is no exaggeration to say that such problems should be solved as soon as possible along with the necessity of stability of the network.
(2) Declaration of attribute data
(a) Reliability of user declaration
As was described above, in the broadband ISDN, before starting the communication, each user should declare to the network complicated and difficult parameters such as peak traffic, mean traffic, burstiness, terminal equipment type, and service quality (QOS: cell discard rate, transmission delay time, and the like) as attribute data. The materialization of the attribute data is still being studied by CCITT at the present time.
It will be very difficult for the ordinary users to correctly understand the meaning of the attribute data and to correctly estimate and declare to the network each parameter value with respect to the call to be made. To prevent that, at the sacrifice of the flexibility, which is the greatest feature of the broadband ISDN, several service items which are combinations of the above mentioned attribute data will be provided so that the users can select them.
Although the network obtains a communication resource in accordance with declared data, each user can unidirectionally send cells to the network regardless of the declared data.
To allow the network, which tries to deliver received cells with the best effort, to stably and effectively operate, the correctness of the declared attribute data is preconditionally required.
Another problem for materializing the broadband ISDN is how to establish with flexibility suitable for various communication needs of the users the declaration method and operation method of attribute data which do not adversely affect the stability and the like of the network even if the declared data contains errors and false data.
(c) Charging method
Another subject to be solved in future is what charging method is set in the broadband ISDN network. At the present time, this subject is not being studied. However, depending on the charging method being set, it will remarkably affect the stable operation of the network and the design of the communication equipment such as switches. Thus, this subject should be solved as soon as possible.
Of course, the charging method to be set should be reflected with attribute data declared by each user. The charging method should be user friendly, false resisting, and inductive of proper use of communication resources in the network. Besides simplification of attribute data, establishment of method with general view is required.
(3) Construction/process of broadband switches
(a) Buffer capacity of ATM switch
As was described earlier, for the ATM switches, each of which is a key element for accomplishing the broadband ISDN, cell buffers with a large storage capacity are required. In addition, each switch should achieve a throughput of 155.52 Mbps or 622.08 Mbps. To break the engineering problems of the switch including the development of the construction method for improving the effectiveness thereof, they are being intensively studied and developed in many laboratories and the like. For each ATM switch employing the Bather-Banyan network method, the common buffer method, or the like, which are considered at the present time, even if it is small in scale such as 8.times.8, a buffer with large capacity for storing several hundred cells will be required as will be described later in more detail. The buffer with large capacity is mainly used when calls with large burstiness are gathered in a path in the same direction. Such a buffer remarkably disturbs the production of large capacity of the ATM switch LSIs along with the suppression of variation of cell delay. By reconsidering the method suitable for the broadband ISDN, if the storage capacity of the buffer were be remarkably reduced, the effect will unexpectedly become large.
(b) Priority control
As was described earlier, the ATM switch determines and controls which cells are output to a desired path with a high precedence in accordance with the service quality (such as cell discard rate and communication delay time) that each user has declared and or whether they are non-priority cells or violation cells. Such a control means prevents the ATM switch LSI from providing large capacity and from decreasing the cost thereof along with necessity of high speed cell buffer with large storage capacity.
(c) Policing function
As was described earlier, at the present time, each user can unilaterally send cells to the network even if the attribute data (or service item) is not what was declared to the network when he or she has set the call. A function for supervising the range of the attribute data (or service item) and for adding it with a violation cell mark if the cell violates the range of the attribute data (or service item) may be provided in each subscriber line interface, which is followed by the ATM switch, so as to further improve the supervisory function, to discard cells in accordance with the traffic condition in the network, and to prevent excessive cells from entering the ATM switch. Thus, the some improvements such as decrease of the storage capacity of the buffer may be expected for the ATM switch. However, only with such improvements, many problems and subject involved in the broadband ISDN network at the present time cannot be comprehensively solved unlike the present invention.
(d) Call acceptance control
A call acceptance control is used to estimate a communication resource necessary in the network in accordance with the attribute data declared by the user and to determine whether or not the resource is accepted. To satisfy a particular traffic intensity necessary for the switch, this control should be accurately and rapidly processed with a simple algorithm. This control method is being evaluated in related study associations.
(e) Charging/traffic totalization
As was described earlier, in the broadband ISDN which has been studied, the cell discard and preference control are performed in the switch. Thus, a difference takes place between information that each user transmitted and that which was transmitted to the receiver side through the network. In addition, the service quality varies for each call and depends on the traffic condition in the network. Thus, when the charging of cells is measured or the traffic in the network is measured for each destination, it is necessary to provide a calculating function in the network (followed by the ATM switch).
However, although a cell which flows in the network has information named a cell header (5 octets) for performing routing control, it contains a virtual path identifier (VPI), a virtual channel identifier (VCI), and the like, which identify the receiver, rather than information for identifying the sender. Thus, it is difficult practically to accomplish the above mentioned object. In other words, in the available method, the number of cells which are sent at the entrance of the network is measured without a provision for checking whether the cells are really sent to the destinations. Thus, the users will not satisfy the network system at the present time.
(4) User-network interface (UNI)
Thus, far, CCITT has intensively studied the fundamental framework of the broadband ISDN. However, the evaluation of the practical user-network interface is a subject to be studied in future. In evaluating the interface, the method where interface is shared by a plurality of terminal equipments, so-called multi-point multi-drop connections, accomplished in narrow band ISDN, will become important, since the interface provides an ultra high speed transmission. In addition, the interface will require compatibility with a transmission speed of 64 kbps, which is the basic speed of the narrow band ISDN so that the broadband ISDN and the narrow band ISDN are mutually connected.